Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Remove unnecessary pad token #2428

Open
wants to merge 4 commits into
base: main
Choose a base branch
from

Conversation

vigneshwaran
Copy link

@vigneshwaran vigneshwaran commented Aug 12, 2023

What does this PR do?

Instead of padding till max_seq_len, use the maximum length of the batch.

I have provided a simple and concise solution.

What issue(s) does this change relate to?

@vigneshwaran vigneshwaran requested a review from a team as a code owner August 12, 2023 07:19
Copy link
Contributor

@dakinggg dakinggg left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the PR! I'll need to double check that this works ok with FSDP and produces the same eval results as before this PR. Left one initial comment related to one of the test failures

batch = {
'input_ids': torch.stack(inputs),
'input_ids': input_ids,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I believe the test error you are getting is because input_ids and labels should not be the same tensor, because labels gets modified to put a -100 when the labels get rolled so they are aligned for the next token objective.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Fix evaluation code to improve performance
3 participants